Hi, I am running a de novo assembly of a plant ~500Mb genome from PacBio reads of 120X coverage with Falcon. It runs for about 5 days now, done ~20% of the 160000 jobs that it has, and the disc space it takes is 2.5T currently. My free disc is up to 6T, meaning that I have 3.5T free. I am trying to estimate the size that I will need. What is the behaviour of Falcon regarding the temp files that it creates and how much free space on a disc it will need...? Is it something that can be calculated?
2 answers
OK, So I am answering to myself just in case someone who reads it will need it. It is abnormal to have these sizes and these numbers of jobs with Falcon trying to assemble a less than 1G genome with a 120X coverage. A Dbsplit option in the cfg file that takes care of the block size at the dealigner stage is very important for the run. In my case I was running it with -s 50 and once changed to -s 400 it reduced the number of jobs to 196!!, runtime to 34h and disc size to 1.3T in total.
Also, just in case someone needs - here are most of the answers to my questions :
Important text for newbies like me assembling de novo for the first time.
Hello alslonik !
Even if I did not used Falcon but Canu, I once had a similar issue while trying to assemble a large genome with PacBio data and a large depth. I think that calculate the memory needed is higly dependant on the parameters you used for your run, but is not that easy to predict.
I'm not sure if Falcon delete temporary files, and if yes, when will he do so. In order you to avoid waiting for several days of assembling, and then abort spaces issue, you could try to read this : https://github.com/marbl/canu/issues/193 , an old post where I could have a lot of feed back from the Canu dev, they helped me understanding why I was producing so much data and how I could lower the volume of produced data (filtering the raw reads for example using a histogram of reads length for example).
In conclusion, buying a disc is not the only solution, you may be able to produce less data by filterings your reads, and playing with some assembly parametrs (like the minimum size for overlapp of 2 reads.) , I'll be glad to do my best to discuss about this !
Hope this was a bit helpfull,
Cheers,
Roxane
Log in to answer this question.
-
Snakemake temp()
written by kstangline •
-
Can you predict the size of a BAM file based on the fasta?
written by James Reeve
-
Canu 2.1.1 Storage issue at correction/1-overlapper
written by arteen.marashi •
-
GWAS on different forms of phenotype variable.
written by biotechnology415 •
-
Sequel II large data set Falcon parameters query
written by rob234king
-
De novo plant genome assembly
written by alslonik
-
Whole genome plant sequencing
written by agata88
-
Tool for PacBio Denovo assembly of diploid plant genome (size 500mb) having around 52x coverage
written by Sudhir Jadhao
-
Abyss Speed-Up Tricks
written by Darked89
-
How To Cluster 454 Reads?
written by Panos
